Reinforcement Learning with Imbalanced Dataset for Data-to-Text Medical Report Generation

Toru Nishino,Keigo Nakamura,Norihisa Nakano,Tomoki Taniguchi,Ryuji Kano,Motoki Taniguchi,Tomoko Ohkuma,Yuki Tagawa,Ryota Ozaki,Yohei Momoki

doi:10.18653/v1/2020.findings-emnlp.202

Abstract

Toru Nishino, Ryota Ozaki, Yohei Momoki, Tomoki Taniguchi, Ryuji Kano, Norihisa Nakano, Yuki Tagawa, Motoki Taniguchi, Tomoko Ohkuma, Keigo Nakamura. Findings of the Association for Computational Linguistics: EMNLP 2020. 2020.

Highlights

Writing medical reports manually from medical images is a time-consuming task for radiologists
We introduce a new reward, Clinical Reconstruction Score (CRS), to quantify how much information the generated reports retain about the input findings
Table-to-Text model achieved the best CRS, so we selected the table-totext model as a text generation module

Summary

Introduction

Writing medical reports manually from medical images is a time-consuming task for radiologists. Radiologists first recognize what findings are included in medical images, such as computed tomography (CT) and X-ray images. Radiologists compose reports that describe the recognized findings correctly without omission. Doctors prefer radiology reports written in natural language. Other types of radiology reports, such as tabular reports, are difficult to understand because of their complexity. The purpose of our work is to build an automated medical report generation system to reduce the workload of radiologists. The medical report generation system should generate correct and concise reports for the input images. Data imbalance may reduce the quality of automatically generated reports. Medical datasets are commonly imbalanced in their finding labels because incidence rates differ among

Objectives

Methods

Results

Conclusion